GPT-Live introduces a ‘full-duplex’ architecture for ChatGPT, enabling a more natural voice conversation, with follow-up questions and pauses managed in real time.
OpenAI has introduced GPT-Live, a new generation of voice models for ChatGPT designed to make spoken exchanges closer to a real conversation. Thanks to a full-duplex architecture, the model can listen and speak simultaneously, react with brief listening signals ("mhmm", "yes"), and remain silent when the user needs a moment of reflection. For questions that require web research or further reasoning, GPT-Live delegates the task to a more powerful model in the background before reintegrating the answer back into the conversation.
Two versions are being rolled out: GPT-Live-1, by default for paid subscribers, and GPT-Live-1 mini for users on the free plan.
Read the full article on the original source
Go to the source →